HP-UX TCP/IP Performance White Paper, March 2008

6 Monitoring Network Performance

There are many tools available on HP-UX for monitoring network statistics. Of these tools, netstat(1)

and lanadmin(1M) are very useful to monitor network statistics especially to identify the following

conditions which may potentially have an impact on network performance:

• Retransmissions

• Out-of-order packets

• Window probes

• Duplicate Acks

• Ack-only packets

• Bad Checksums

• Dropped Fragments

• UDP socket buffer overflows

• Inbound packet discards/errors

• Outbound packet discards/errors

• Packet Collisions

6.1 Monitoring network statistics

The netstat command displays statistics for network interfaces and protocols. The output format varies

according to the options selected.

The netstat command without any options displays all active internet connections and all active UNIX

domain sockets. This is useful for tracking what connections are being used on the system at any point in

time. Using the -a option shows the state of all sockets, active or not, this option will be useful since one

can observe all the listen sockets too.

The netstat -i command shows statistics for all interfaces. It shows MTU size for each network interface,

and number of inbound and outbound packets processed.

The netstat -s command provides statistics for all the protocols like TCP, UDP, IPv4, ICMP, IGMP, IPv6

and ICMPv6.

To diagnose networking problems, it is best to use data collected over an interval and not all the statistics

since boot. There is an unofficial tool called beforeafter which can be used to subtract one set of

netstat statistics from another. This tool can be retrieved via the URL:

ftp://ftp.cup.hp.com/dist/networking/tools/beforeafter.c

Typically, one might use beforeafter as follows:

# netstat -s > before

# do something or sleep for a time

# netstat -s > after

# beforeafter before after > delta

# more delta